Release 0.14.0

[github-actions]

Release readiness review (v0.13.6 -> TARGET 325db84)

This is a release readiness report done by $final-release-review skill.

Diff

v0.13.6...325db84

Release call:

🟢 GREEN LIGHT TO SHIP Version is bumped to 0.14.0, scope is large but paired with substantial new coverage and no concrete release-blocking regressions found.

Scope summary:

• 463 files changed (+95,304/-1,687); key areas touched: new src/agents/sandbox/** runtime and provider adapters, run.py/run_state.py orchestration and persistence updates, extensive tests/sandbox/** and tests/extensions/test_sandbox_*.py, plus major docs/examples additions.

Risk assessment (ordered by impact):

1. Core dependency and optional dependency shifts

• Risk: 🟢 LOW. Dependency resolver behavior may vary in downstream environments due to base/extra dependency changes.
• Evidence: pyproject.toml adds websockets>=15,<16 to core dependencies, updates litellm extra constraint, and adds many new extras (docker, blaxel, daytona, e2b, modal, runloop, vercel, temporal, s3).
• Files: pyproject.toml, uv.lock
• Action: Run uv sync --all-extras --all-packages --group dev and make tests; pass criteria: dependency resolution succeeds and full test suite passes in CI.

2. Run-state/schema and resume-flow expansion

• Risk: 🟢 LOW. Resume/persistence paths have expanded complexity but include explicit schema/version handling and targeted tests.
• Evidence: CURRENT_SCHEMA_VERSION moved to 1.9 with added summaries and broader serialization/deserialization paths (agent identity keys, sandbox payload, custom tool calls); corresponding tests added/updated under run-state and sandbox suites.
• Files: src/agents/run_state.py, src/agents/run.py, tests/test_run_state.py, tests/sandbox/test_session_state_roundtrip.py, tests/sandbox/test_runtime.py
• Action: Run uv run pytest -q tests/test_run_state.py tests/test_run_impl_resume_paths.py tests/sandbox/test_session_state_roundtrip.py; pass criteria: all targeted persistence/resume tests pass.

Notes:

• Working tree was clean (git status --short returned no changes).
• Base tag was selected from local tags only (v0.13.6) per request.
• Local verification was not re-run in this review; report assumes CI verification for the target commit.